fpls-12-639368 April 30, 2021 Time: 12:6 # 1

ORIGINAL RESEARCH published: 30 April 2021 doi: 10.3389/fpls.2021.639368

Cycad-Weevil Pollination Symbiosis Is Characterized by Rapidly Evolving and Highly Specific Plant-Insect Chemical Communication

Shayla Salzman1,2*, Damon Crook3, Michael Calonje4, Dennis W. Stevenson5, Naomi E. Pierce2† and Robin Hopkins2†

1 Plant Sciences, Cornell University, Ithaca, NY, United States, 2 Organismic and Evolutionary Biology, Harvard University, Cambridge, MA, United States, 3 Otis Laboratory, USDA-APHIS-PPQ CPHST, Otis ANGB, MA, United States, 4 Montgomery Botanical Center, Coral Gables, FL, United States, 5 New York Botanical Garden, Bronx, NY, United States

Coevolution between plants and insects is thought to be responsible for generating biodiversity. Extensive research has focused largely on antagonistic herbivorous

Edited by: relationships, but mutualistic pollination systems also likely contribute to diversification. Dani Lucas-Barbosa, Here we describe an example of chemically-mediated mutualistic species interactions ETH Zürich, Switzerland affecting trait evolution and lineage diversification. We show that volatile compounds Reviewed by: produced by closely related species of Zamia cycads are more strikingly different Nina E. Fatouros, Wageningen University and Research, from each other than are other phenotypic characters, and that two distantly related Netherlands pollinating weevil species have specialized responses only to volatiles from their Sergio E. Ramos, University of Pittsburgh, United States specific host Zamia species. Plant transcriptomes show that approximately a fifth *Correspondence: of genes related to volatile production are evolving under positive selection, but Shayla Salzman we find no differences in the relative proportion of genes under positive selection [email protected] in different categories. The importance of phenotypic divergence coupled with † These authors have contributed chemical communication for the maintenance of this obligate mutualism highlights equally to this work chemical signaling as a key mechanism of coevolution between cycads and their Specialty section: weevil pollinators. This article was submitted to Plant Metabolism Keywords: coevolution, brood-site mutualism, chemical signaling, Coleoptera, Caribbean and Chemodiversity, a section of the journal Frontiers in Plant Science INTRODUCTION Received: 08 December 2020 Accepted: 12 April 2021 Obligate mutualistic associations involving a high degree of specialization by both partners such Published: 30 April 2021 as those between figs and fig wasps and yuccas and yucca moths are less common in nature, Citation: and yet the analysis of such systems is also vital to our understanding of mechanisms underlying Salzman S, Crook D, Calonje M, the evolution of plant/insect interactions (Pellmyr, 2003; Herre et al., 2008). Studies of these Stevenson DW, Pierce NE and and other pollination systems (Holland and Fleming, 1999; Kawakita, 2010) have highlighted Hopkins R (2021) Cycad-Weevil morphological and behavioral traits associated with mutualism, and begun to identify mechanisms Pollination Symbiosis Is Characterized of diversification and speciation associated with these lineages (Althoff et al., 2012; Hembry by Rapidly Evolving and Highly Specific Plant-Insect Chemical et al., 2013; Suinyuy et al., 2015). Plant volatile production for attracting pollinators has been Communication. hypothesized as a likely mediator of specialized brood-site mutualisms (Pellmyr, 2003; Okamoto Front. Plant Sci. 12:639368. et al., 2007), yet has been largely overlooked in its contribution to diversification (Suinyuy et al., doi: 10.3389/fpls.2021.639368 2015). In order to fully understand the process of diversification in mutualistic, coevolving lineages,

Frontiers in Plant Science| www.frontiersin.org 1 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 2

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

we need to learn more about the causes and consequences of Caribbean Zamia clade to look for evidence of rapid evolution species interactions. Specifically, we are interested in identifying: and positive selection associated with volatile production. By (1) traits driving these relationships, (2) whether these traits analyzing reciprocal evolutionary change in volatile production are diverging across closely related taxa, and (3) the role of and insect perception, we provide investigation of coevolution selection in trait divergence. Zamia cycads and their Rhopalotria in what is arguably the earliest example of an insect mediated pollination mutualists are an ideal study system as they represent “pollination syndrome” (Cai et al., 2018; Salzman et al., 2020). an understudied obligate brood-site pollination mutualism where the importance of volatile signaling is clear. The ancient plant order Cycadales is an early diverging MATERIALS AND METHODS lineage of gymnosperms (Ran et al., 2018) and, unlike most other gymnosperms requires insect pollination instead of wind Study System and Design pollination (Terry et al., 2012). The pollination mechanism of Rhopalotria and Zamia form a tight obligate mutualism. most cycads appears to be an obligate brood site mutualism Rhopalotria and their close relatives are found across much of whereby highly specialized pollinators live out their lifecycle Zamia diversity where they provide necessary pollination services within pollen cone tissue (Mound and Terry, 2001; Terry, 2001; to their respective host Zamia species, which in turn provide Terry et al., 2005, 2007). This mutualism is hypothesized to be food, brood sites, and shelter. Feeding damage does not affect driven by host plant volatile organic compounds (VOCs) through reproduction or potential fitness of Zamia as Rhopalotria feed a push-pull pollination mechanism (Terry et al., 2007; Salzman only on the disposable pollen cone parenchyma tissue. This et al., 2020). Pollinators are attracted to mid-level amounts of the mutualism is mediated by plant volatile production (Salzman host plant compound but are repelled by high levels of VOCs. et al., 2020), a cue that is important for lifecycle completion Pollen and ovulate cones have a daily cycle of VOC production in both lineages. The strength of the mutualism is such that causing pollinators to be repulsed by the male cones and attracted pollinators go into diapause during the 10 months of the year to the female cones, and then attracted to the male cones once when reproductive services are not required by the host plant again. The pattern of plant volatile release is conserved across (Norstog and Fawcett, 1989; Norstog et al., 1992). the Cycadales, and the behavioral response of pollinators to the Caribbean Zamia species are distributed across Florida, Cuba, daily change in expression of host plant VOC has converged Jamaica, the Bahamas, Cayman Islands, Dominican Republic (but between Cycadothrips chadwicki (Thysanoptera) and Rhopalotria not Haiti) and Puerto Rico (Calonje et al., 2020). Rhopalotria furfuracea (Coleoptera) suggesting that this push-pull pollination species have been found associated with all populations of mechanism is the likely pollination syndrome across the Caribbean Zamia except those on Puerto Rico (Tang et al., Cycadales (Salzman et al., 2020). Much of the Neotropical 2018). Zamia integrifolia as currently circumscribed is distributed cycad genus Zamia are pollinated by Rhopalotria weevils and across the southeastern United States (Florida and Georgia), closely related genera within the Allocorynina (Tang et al., 2018). the Bahamas, Cayman Islands, and Cuba (Calonje et al., 2020) These weevils feed and reproduce on the pollen cone of their although these populations are not monophyletic (Calonje et al., host Zamia (Norstog and Fawcett, 1989; Norstog et al., 1992). 2019). Rhopalotria slossoni is distributed across southern Florida Conversely, the host Zamia species are completely dependent (O’Brien and Tang, 2015) where it provides pollination services upon pollination services by the weevils (Norstog et al., 1986; to populations of Z. integrifolia (Tang, 1987; Norstog et al., 1992). Tang, 1987). The Zamia—Rhopalotria symbiosis appears to be fully mutualistic for both lineages, and the relationship has Plant Volatile Analysis been hypothesized to exhibit phylogenetic congruence between To quantify the diversity of VOC profiles in the Zamia clade, specialized partners (Norstog et al., 1992; Stevenson et al., 1998; we collected plant volatiles from dehiscing pollen cones of 10 Tang et al., 2018). However, studies remain to be done that Caribbean taxa using headspace collection methods (see Salzman provide information about potential coevolutionary mechanisms et al., 2020). We quantified volatiles from 3 to 7 plants per taxa. involved, including those documenting reciprocal evolutionary These plants were wild collected as seeds from native populations change in each lineage (e.g., Janzen, 1980). across the Caribbean (Supplementary Figure 1) and cultivated Chemical communication mediating the relationship between in a common garden at Montgomery Botanical Center (MBC) plant and pollinator has been described for one species pair, in Coral Gables, Florida. All volatiles were collected between Mexican Z. furfuracea/R. furfuracea, and is hypothesized to be the hours of 13:30 and 16:00 (Supplementary Table 1) across 7 driving relationships across the two lineages (Salzman et al., days in 2019. Zamia phenology across the phylogeny is highly 2020), but we do not know whether the plant/pollinator behavior varied at MBC, but most Caribbean Zamia are reproductive in is shared across other Zamia/Rhopalotria pairs, whether volatile January/February (Griffith et al., 2012). We also captured volatiles production and perception traits are diverging in the lineages, from one pollen and two ovulate cones of Z. integrifolia collected or whether any of these traits are driven by positive selection. from Levy County, Florida at hour and a half time points from Here we describe the chemical communication that underlies 8:00 until 20:45. MBC accession numbers for these plants are a distantly related species pair, the Caribbean Z. integrifolia— 20050825∗A, 20050831∗A, and 20050820∗C. Samples were eluted R. slossoni. We then ask if there has been evolution of specificity as described in Salzman et al.(2020) and were run on a combined in volatile perception in these two Rhopalotria species. Finally, Agilent technologies 6890 network gas chromotograph and 5973 we investigate phenotypic and genotypic evolution across the mass-selective detector using a DB-5 column (J&W Scientific

Frontiers in Plant Science| www.frontiersin.org 2 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 3

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

Inc., Folsom CA; 30 m × 0.25 mm I.D.; film thickness, 0.25 um; MO, catalog no. M2047). For these standard GC-EAD runs, the splitless mode) (GC-MS). Helium was the carrier gas at a constant oven temperature was held at 50◦C for 1 min, and then increased flow rate of 0.7 ml/min. the injector temperature was 250◦C, oven 20◦C /min to 320◦C, where it was held for 5 min. All other temperature was held at 50◦C for 1 min, and then increased 10◦C settings remained the same. /min to 170◦C where it was held for 5 min. To determine the specificity of weevil perception of potential Volatile peaks were standardized and calibrated manually host plants, the GC-EAD physiological responses of two species to include all peaks to be comparable across samples. Bag, of Rhopalotria were compared using the volatile compounds of filter, and DCM controls showed a few contaminant peaks after their host and non-host Zamia species. The active compound 15 min retention time that were removed during integration. The identified using antennae of R. slossoni, methyl salicylate, differs ChemStation Integrator was used with the following parameters: from the active compound identified in another Rhopalotria- Initial area reject = 0, initial peak width = 0.300, initial Zamia species pair involving Zamia furfuracea. In the latter, threshold = 16.0, shoulder retention = off. Integration began the plant volatile compound 1,3-octadiene was identified as at 3,550 min after the solvent peak and ended at 15,000 min the physiologically active component eliciting a response from (ChemStation software, Agilent technologies). Peak areas were the mutualistic weevil partner, R. furfuracea (Salzman et al., calculated to percent composition (size of individual volatile 2020). Thus R. slossoni (n = 2) and R. furfuracea (n = 3) compound peak/total peak area of the sample) and were treated were tested against a mixture of 1,3-octadiene (ChemSampco as phenotypic traits in subsequent trait analysis. Inc. Dallas, TX, catalog no. 7015.90) and methyl salicylate Percent compositions of volatile peaks were analyzed (Sigma-Aldrich, St. Louis, MO, catalog no. M2047) using the for dissimilarities within and between species (Bray-Curtis same GC-EAD set up and oven settings used for the cone dissimilarity index with 999 permutations) using the non- aeration GC-EADs. parametric test ANOSIM (analysis of similarities), vegan v2.3-5 package (Oksanen et al., 2019) in R version 3.6.3 Weevil Behavior (R Core Team, 2020). Using pitfall tests, we confirmed that the behavioral response of R. slossoni to the volatile compound methyl salicylate follows push-pull pollination in the same manner observed Weevil Electroantenograph Detection in R. furfuracea (see Salzman et al., 2020). To determine (EAD) whether methyl salicylate acts as an attractant, and whether Gas Chromotrography-Electroantenograph Detection (GC- weevil attraction changes with differing amounts of the volatile EAD) can be used to determine the physiological capability of an compound, behavioral assays were carried out consisting of a insect to perceive a volatile compound. We used this technique hexane control and five dilutions of methyl salicylate in HPLC to identify volatile compounds produced by Zamia integrifolia grade hexane (concentrations: 1 ng/µl, 10 ng/µl, 100 ng/µl, 1 that trigger responses in the antennae of Rhopalotria slossoni. We µg/µl, 10 µg/µl) (see Salzman et al., 2020 for detailed methods). also tested the specificity of pollinator response using two species 10 µl of each dilution was used to cover the natural emission of of Rhopalotria and their respective host plants. methyl salicylate production found in Z. integrifolia (Figure 1C). We used headspace collection methods (see Salzman et al., Methyl salicylate values at peak plant expression ranging from 1 2020) to capture volatiles from receptive ovulate (Montgomery to 50 µg (Figure 1C) are expected to be repellent to R. slossoni. Botanical Center (MBC) accession number 20050831∗A) and All concentrations plus the control were run simultaneously for dehiscent pollen (MBC accession numbers 20050815∗1 and each trial (n = 10). Prior to the trial, weevils were allowed to 20050815∗2) cones of Z. integrifolia. We then used GC-EAD (see feed freely on Z. integrifolia cones so they would not be stressed Crook et al., 2008; Salzman et al., 2020) to determine which host before the trial started. Each trial was carried out by placing plant volatiles elicited responses from the antennae of R. slossoni. 4 R. slossoni weevils into the arena away from the pit. Arenas We injected 2 µl plant volatile eluate and used a GC method were closed and the entire trial was placed in the dark at room where the oven temperature was held at 50◦C for 1 min, and temperature (21◦C) for 30 min, after which time the number then increased 10◦C/min to 300◦C where it was held for 5 min. of weevils in the pits were counted. Dead or copulating weevils The injector temperature was 280◦C and the GC outlets for were not counted. the EAD and FID were 300◦C. For mass spec identification of As with R. furfuracea (Salzman et al., 2020), R. slossoni individual plant volatile compounds, samples were run separately behavioral response to host volatile did not follow a normal as described above for plant volatile analysis. distribution and therefore the raw data plus 0.5 (to account for Eleven R. slossoni weevils were tested using GC-EAD against zero values) was log transformed for statistical tests of differences Z. integrifolia cone aerations. The antennally active peak was in attraction to the different concentrations of methyl salicylate. identified on the basis of its mass spectra using the universal NIST Weevil attraction to different amounts of methyl salicylate was chemical library (NIST version 2.0, 2002), Kovats index (van determined through an ordinary least squares model when no Den Dool and Kratz, 1963; Kovats, 1965), and by comparing its significant difference was found between ordinary least squares retention time and mass spec with an authentic standard obtained and weighted least squares models (p = 0.54) using the nlme from Sigma-Aldrich (St. Louis, MO, catalog no. M2047). The package (Pinheiro et al., 2020) in R version 3.6.3 (R Core Team, positive EAD response was tested using GC-EAD of nine weevils 2020). P-values were then corrected for family-wise error using against a standard of methyl salicylate (Sigma-Aldrich, St. Louis, sequential Bonferroni correction.

Frontiers in Plant Science| www.frontiersin.org 3 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 4

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

FIGURE 1 | Continued Volatiles were sampled from 8:00 to 20:45. Peak emissions are between 13:30 and 16:00 h. Methyl salicylate production in pollen cones during off peak hours (inset) coincides with the amount most attractive to weevils in pitfall tests. Production of methyl salicylate during peak hours coincides with amounts found to be less attractive in pitfall tests. Ovulate cones also increase their emission of methyl salicylate, reaching peak attractive amounts at the same time that pollen cones become less attractive.

Morphological Trait Analysis To quantify morphological variation across the Zamia clade, we photographed Caribbean Zamia herbarium sheets from 12 Herbaria and processed them for morphological leaflet measurements in ImageJ (version 2.0.0-rc-69/1.52n) (Rueden et al., 2017) through Fiji (Schindelin et al., 2012). Leaflet length/width were chosen as comparative morphological traits because they have been shown to be fairly consistent within and divergent between Caribbean populations (Eckenwalder, 1980). Herbarium sheets were used from A, ANS-PHILA, BNH, FTG, GH, MAPR, MICH, MO, NYBG, UPRRP, US, and USNH herbaria. We also collected leaflets from natural populations on Eleuthera, Andros Island, Grand Bahamas, New Providence and Long Island. Leaflets from the middle of the leaf rachis were selected from each herbarium sheet or individual field collection and measured for length and width. Cycad leaflets unfurl and all leaflets expand at the same time resulting in a leaflet shape that is mostly uniform across the leaf while there may be variation in size especially at the top and bottom of the leaf rachis. Size increases as the leaflet grows and fully expands. Due to the limitations of the available material and not knowing the age or developmental stage of the leaflets in herbarium material, we calculated ratios of leaflet length by leaflet width to remove any influence of size. The number of measured leaflets per population are: Zamia sp. (Andros Island): 34, Z. angustifolia (Eleuthera): 10, Z. sp. (Eleuthera) 48, Z. sp. (Grand Bahamas): 51, Z. sp. (New Providence): 37, Z. lucayana (Long Island): 25, Z. integrifolia (Florida): 19, Z. pumila (Puerto Rico): 12, Z. erosa (Puerto Rico): 6. GPS point localities were either measured in the field, copied from herbarium sheets, or manually generated using herbarium data sheet locality information and Google Maps to determine their inclusion in the studied populations. Herbarium sheets were not available for the Jamaican populations of Z. aff. amblyphyllidia or Z. aff. portoricensis and so these species were not included in the leaflet morphology analysis. We analyzed the leaflet measurements using R version 3.6.3 (R Core Team, 2020) to determine dissimilarities (Bray-Curtis FIGURE 1 | Rhopalotria slossoni perception and behavioral response to dissimilarity index with 999 permutations) using within and Zamia integrifolia production of methyl salicylate follows push-pull pollination (Salzman et al., 2020). (A) Electroantenograph detection (EAD) shows methyl between populations using the non-parametric test ANOSIM salicylate elicits responses from antennae of R. slossoni. Plant gas (analysis of similarities) as implemented in the vegan v2.3-5 chromatography flame ion detection (GC-FID) on the top, weevil package (Oksanen et al., 2019). electroantenograph detection (EAD) on the bottom with red arrow denoting a positive response (n = 11 and confirmed with a standard n = 9). (B) Pitfall test of behavior shows that methyl salicylate acts as an attractant with a non-linear Selection in the Zamia Genome response. Median values are shown in red and p-values are in Transcriptomes Supplementary Table 5. (C) Methyl salicylate emissions from Zamia We collected Plant RNA from dehiscing pollen cones of integrifolia pollen and ovulate cones change over the course of the day. (Continued) wild collected Caribbean Zamia seeds grown at MBC for three purposes: (1) to detect signatures of positive selection

Frontiers in Plant Science| www.frontiersin.org 4 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 5

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

across the Caribbean clade, (2) to identify genes whose Van Dongen and Abreu-Goodger, 2012) on all non-redundant expression correlates with volatile production, and (3) to identify coding sequences to obtain clusters of similar sequences. These genes whose expression correlates with cone development. clusters are aligned using MAFFT v7.309 (Katoh and Standley, To detect signatures of selection across the Caribbean clade, 2013) and trimmed with Phyutility (Smith and Dunn, 2008). microsporophyll (cone scales) were collected from dehiscing Cluster trees were estimated using FASTTREE v2.1.9 (Price et al., pollen cones of 10 Caribbean Zamia (Supplementary Table 2). 2010). In order to account for long branches resulting from To identify genes whose expression correlates with volatile misassembly, paralogy, or recombination branches longer than production, microsporophyll cone scales were collected from 10 times the average distance to tips in its sister clade and longer Z. integrifolia at six time points across a day in concert with than 0.4 substitutions per site were trimmed from these trees. volatile headspace sampling. This sampling was done at the same The resulting sequence files were then realigned using MAFFT time as the daily volatile collection described above using MBC (Katoh and Standley, 2013) and used to infer gene trees with accession 20050825∗A. Volatiles were collected for 45 min, at Randomized Axelerated Maximum Likelihood (RAxML) v8.2.10 which point one microsporophyll cone scale was removed from (Stamatakis, 2014). These trees were subjected to an additional the center of the cone for RNA extraction. To identify genes long branch trimming and subsetted to a taxon occupancy of 12 whose expression correlates with cone development, entire cones or more for phylogenetic analysis and 10 or more for tests of of Z. furfuracea were collected at three developmental stages: just positive selection that excluded the out-group, Zama furfuracea, emerging cones, young half sized cones, and immature almost and the Caribbean Z. lucayana where many genes were missing full-sized cones. Two separate plants were sampled, with one cone due to poor transcriptome assembly. These homologous gene of each stage collected on each plant. trees were then further pruned to a single orthologous sequence All RNA samples were flash frozen in liquid nitrogen in the per sample using the maximum inclusion method (MI) which field and then stored at −80 until RNA extraction. Accession isolates the subtree with the highest number of taxa without information and initial collection locality for all plants are taxon duplication (Dunn et al., 2008, 2013; Smith et al., 2011; listed in Supplementary Table 2. Cone tissue was ground Yang and Smith, 2014). to a powder in liquid nitrogen and RNA extraction was performed using the Qiagen RNeasy Plant MiniKit (Qiagen, Plant Genomic Positive Selection Hilden, Germany) with the addition of 100 µl polyethylene Transcriptomes of dehiscing pollen cones of ten Caribbean glycol 4000 and 10 µg polyvinlypyrrolidone 40. RNA purity Zamia species were used to determine whether genes related to and integrity were evaluated using a 2100 bioanalyzer (Agilent volatile production were experiencing higher rates of positive Technologies, Santa Clara CA) using the plant RNA Nano Assay selection than other genes in the genome. The Mexican Zamia (version 1.3). Libraries were prepared from samples with a furfuracea was included as an outgroup. This was done in three RNA Integrity Number greater than 7 using the NEBNext Ultra steps briefly described here and further described below. First, RNA library prep kit for Illumina (# E75305, New England genes that were found in all in-group species transcriptomes Biolabs, Ipswich MA), NEBNext Poly(A) mRNA magnetic were tested for evidence positive selection. Second and separately, isolation module (#E7490, New England Biolabs, Ipswich MA), differential expression analysis was run on Z. furfuracea NEBNext multiplex oligos for Illumina (#E7335, #E7500, New developmental cones to identify differentially expressed (DE) England Biolabs, Ipswich MA), and Agencourt AMPure XP genes related to cone development and on Z. integrifolia cone beads (#A63881, Beckman Coulter, Pasadena CA). Library purity scales collected in correlation with volatile production to identify and Integrity was assessed using a 2100 bioanalyzer (Agilent DE genes related to volatile production. These two groups of Technologies, Santa Clara CA) using the high sensitivity DNA DE genes were then used to define three gene “bins”: DE Assay (version 1.03). Libraries averaged 350–500 base pairs. All genes related to Z. furfuracea cone development = “reproductive 23 samples were combined in equal concentration and run across development genes,” DE genes related to Z. integrifolia volatile three lanes of 125 paired end reads on Illumina HiSeq2500 (San production = “volatile associated genes,” and all remaining Diego CA) at Cold Spring Harbor Labs (Cold Spring Harbor NY). genes = “other genes.” Finally, these two data sets were brought Transcriptomes were filtered to remove low-quality paired- together and the genes previously found to be experiencing end sequence reads with Timmomatic (Bolger et al., 2014) and positive selection were divided amongst gene “bins” to look for individually de novo assembled using Trinity v2.3.2 (Grabherr increased positive selection on volatile associated genes. et al., 2011; Haas et al., 2013; Supplementary Table 3). First, we estimated a species tree for all Caribbean Zamia Assembly quality was assessed using the embroyphyta lineage included in this study using 829 orthologous genes found in all 11 gene set in BUSCO v3.0.2. (Seppey et al., 2019). Through the in-group samples and the Zamia furfuracea out-group. Orthologs Galaxy platform (Afgan et al., 2018; Supplementary Figure 2). were concatenated for a total of 797,745 aligned columns with an Coding sequences were identified with Transdecoder v3.0.0 overall matrix occupancy of 0.96 (Statistics per sample are found (Haas et al., 2013) and sequence redundancy reduced with in Supplementary Table 4). Models of evolution for each gene CD-HIT-EST v4.6.4 (Li and Godzik, 2006). Orthologs were were determined using ModelFinder plus in IQ-TREE v 1.6.1 identified using the tree based ortholog identification pipeline (Nguyen et al., 2015; Kalyaanamoorthy et al., 2017). All partitions described in Yang and Smith(2014) and briefly described here. were set to share the same branch lengths, with each partition This involved using all-by-all BlastN (Altschul et al., 1997) allowed to have its own evolution rate. A maximum likelihood and Markov cluster algorithm (MCL) (Enright et al., 2002; (ML) tree and bootstrap tree were then estimated together using

Frontiers in Plant Science| www.frontiersin.org 5 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 6

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

200 runs and random bootstrap and parsimony seeds in RAxML Yang et al., 2000; Wong et al., 2004) and computed p-values v8.2.10 (Stamatakis, 2014; Figure 2A). according to a χ2 distribution with two degrees of freedom. We then investigated evidence of positive selection on the A false discovery rate (FDR) was applied to correct p-values for 2,487 genes that were found in 10 in-group Caribbean species multiple testing using the fdrtool R package (Strimmer, 2008). using the site models (Nielsen and Yang, 1998; Yang et al., 466 genes were statistically significant at 95% confidence for the 2000) implemented in the program Phylogenetic Analysis by preferred M2a model of selection verses the M1a neutral model Maximum Likelihood v4.8 (PAML) (Yang, 2007). The species tree and 463 genes were statistically significant for the preferred was pruned to remove the two taxa not used in PAML selection M8 model of selection verses the M7 neutral model. The 463 analysis, the out-group Z. furfuracea and in-group Z. lucayana genes where both comparisons favored the selection model were that had a poor quality assembly as determined with BUSCO considered to be those genes expressed in the dehiscing pollen scores. Nucleotide sequences for these genes were aligned using cone and shared across Caribbean Zamia that are experiencing the codon model in prank v.170427 (Löytynoja and Goldman, positive selection. 2008). Individual PAML analyses for each of the 2,487 genes were Separately, we used gene expression patterns to create run using scripts from Schultz and Sackton(2019). In tests of three gene groupings: “volatile associated genes,” “reproductive positive selection, ω (the ratio of non-synonymous/synonymous development associated genes,” and “other genes.” For the mutations) is determined for sites along an alignment. We tested “volatile associated genes,” we determined gene expression four site models that allowed for the ω ratio to vary among correlated with volatile production using Z. integrifolia samples codons or amino acids in the protein. These models were M1a that were collected in concert with volatile collection. For the (neutral), M2a (selection), M7 (beta), and M8 (beta and ω). We “reproductive development associated genes” we determined then computed likelihood ratio tests comparing likelihood scores gene expression correlated with pollen cone development using between M1a vs. M2a and M7 vs. M8 (Nielsen and Yang, 1998; Z. furfuracea samples. For both Z. integrifolia volatile and

FIGURE 2 | Cone volatile diversity is more different between Caribbean Zamia species than leaflet morphology. (A) Maximum likelihood phylogeny using 829 orthologous genes of Caribbean Zamia showing leaflet and volatile phenotypic diversity. Bootstrap values are placed at the nodes. The out-group Z. furfuracea has been removed from the image for clarity. Representative leaflets are included to scale for each population where living collection or herbarium records were available and a schematic of cone volatiles is provided. Individual volatile compounds are represented by boxes along the schematic and are aligned between species. Volatiles are presented as present or absent (colored or white box) and percent composition is averaged across all individuals for the species (intensity of color). Methyl salicylate is denoted with a red arrow. A close-up of these aligned volatile representations with corresponding GC-MS retention times is presented in Supplementary Figure 3, along with a dendrogram heatmap of each sample in Supplementary Figure 4. (B) Analysis of similarity (ANOSIM) results using leaflet shape (length/width) measurements taken from herbarium and living collections (top) and percent compositions of all volatile compounds (bottom). Boxplots of ranked dissimilarities are shown for each taxon as well as between taxa and widths are proportional to sample size. The higher the R-value, the more dissimilar samples are between species than within. Species without data are not included in the analysis.

Frontiers in Plant Science| www.frontiersin.org 6 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 7

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

Z. furfuracea development, expression for each individual sample bin were annotated using Trinnotate v3.2.1 (Bryant et al., 2017). was calculated using RSEM v1.2.29 (Li and Dewey, 2011) and The 21 identified gene ontology molecular functions (Ashburner the resulting count matrices were analyzed using the EBSeq et al., 2000; Gene Ontology Consortium, 2021) were then used package v1.28.0 (Leng et al., 2013; Leng and Kendziorski, to categorize genes into 7 functional categories (Supplementary 2020) in R v4.0.3 for conditional analysis. To determine gene Figure 6). expression pattern conditions for the “volatile associated genes,” we used the total ion abundance of methyl salicylate present in volatile samples collected in concert with Z. integrifolia RESULTS assembled transcriptomes. Six time points from one individual were analyzed and found varying expression of methyl salicylate We investigate the coevolution of the Rhopalotria and Zamia (Supplementary Figure 5A). However, time points one and lineages by determining that the mechanism of interaction two had similar total ion abundances (79,531 and 83,357, is consistent across two species pairs and provide evidence respectively) as did time points three and five (46,720 and of reciprocal evolution in traits related to this mechanism 41,758, respectively). Therefore, gene expression patterns that in both lineages. Using insect physiology and behavior, we were considered correlated with volatile production were the demonstrate that plant-insect chemical communication and three in which these time point pairs, one and two and three push-pull pollination is indeed consistent across these two species and five, showed similar expression (Supplementary Figure 5A). pairs and is driving the relationship of the Rhopalotria/Zamia Additionally, two gene expression patterns considered non- mutualism. We then show that these traits are diverging in both volatile associated were included, those where expression patterns lineages and that there appear to be higher rates of positive were all the same or all different across all time points. selection associated with volatile production. To determine gene expression pattern conditions for the “reproductive development associated gene,” we used three Chemical Communication Is a developmental stages of Z. furfuracea pollen cones: just emerged, young half sized cones and immature almost full-size cones. Mechanism of Mutualism Across Gene expression patterns that were considered correlated with Rhopalotria and Zamia Species Pairs pollen cone growth were those that differed by developmental We identify the chemical communication underlying the stage as either all stages different or the first or last stage R. slossoni/Z. integrifolia mutualism and find the push-pull different (Supplementary Figure 5B). The expression pattern mechanism to match that demonstrated in R. furfuracea/Z. with no difference between reproductive stages was included and furfuracea (Salzman et al., 2020) in which weevils have specialized considered not associated with reproductive structure growth. to perceive only the signal emitted by their respective host plant. Gene expression conditions were determined using EBSeq as To do this, we first identified plant volatile compounds that are libraries were set to vary between samples and 25 iterations were physiologically perceived by the pollinator in the R. slossoni/Z. run on both datasets separately until convergence was reached integrifolia species pair. Using electroantenograph detection as determined by a difference of less than 0.001 in the hyper- (EAD) we find that R. slossoni is physiologically capable of parameters α and β and the mixture parameter ρ for each run. perceiving only one compound emitted by its host plant, and Genes were considered volatile associated if they had a greater we identified this as methyl salicylate (Figure 1A), a compound than 0.95 posterior probability of being in one of the three volatile previously reported from Z. integrifolia (Pellmyr et al., 1991). associated gene expression patterns (Supplementary Figure 5A). In order to begin to determine whether chemical Genes were considered reproductive development associated communication and push-pull pollination is indeed the genes if they had a posterior probability greater than 0.95 of being mechanism of interaction across Rhopalotria/Zamia lineages, in one of the three reproductive stages gene expression patterns we compared the volatile behavioral response and plant volatile (Supplementary Figure 5B). production in R. slossoni/Z. integrifolia with those identified The volatile and reproductive associated genes were then in an earlier study of Rhopalotria/Zamia mutualism. Using filtered to include only those genes that were present in the the same pitfall tests we used in the earlier work, we found 2,487 genes used in the positive selection PAML analysis. that R. slossoni are attracted to methyl salicylate but that the Additionally, any genes that were found in both the volatile and response is non-linear (Figure 1B). Weevils are significantly the reproductive associated gene bins were removed from both, more attracted to mid-level amounts of methyl salicylate than generating the final “volatile associated” bin with 146 genes and to low or high amounts (p-values in Supplementary Table 5). “reproductive development associated” bin with 674 genes. The To determine if methyl salicylate could be performing the “other” gene bin includes all remaining 1,667 genes not found push-pull pollination function in Z. integrifolia, we analyzed the in either the “volatile associated” and “reproductive development production of methyl salicylate across the day. We found that associated” gene bins as well as those genes that were found methyl salicylate production does undergo a large increase in the to overlap. Finally, we determined the distribution of positive pollen cone (Figure 1C). selection in Caribbean Zamia pollen cone transcriptomes across To test whether insect perception of host plant compounds gene bins by blasting the 463 genes found to be experiencing may be diverging across these lineages, we performed a reciprocal positive selection against the three gene bins (Supplementary test of volatile perception between R. slossoni and R. furfuracea, Figure 5C). The 30 positively selected genes in the “volatile” an obligate pollinator of Z. furfuracea previously found to

Frontiers in Plant Science| www.frontiersin.org 7 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 8

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

perceive and respond to the Z. furfuracea compound 1,3- genes, of which we measured 463 (18.6%) experiencing positive octadiene (Salzman et al., 2020). Using mixtures of methyl selection. Our differential expression analysis and filtering salicylate and 1,3-octadiene, we find that weevils show a positive identified 146 “volatile associated” genes, 674 “reproductive electroantenographic response to the compound produced by development associated” genes, and 1,667 “other” genes among their host plant, but that neither weevil responds to the the 2,487 orthologs. The ratio of positively selected genes among electroantennally active compound of the other. the “volatile associated” genes was 30 out of 146 (20.6%), among the “reproductive development associated” genes was 144 out Zamia Volatile Phenotypes Are More of 674 (21.4%), and among “other” genes was 289 out of 1,667 Diverged Than Morphological (17.3%) (Supplementary Figure 5). Although it is tempting to Phenotypes interpret this as showing a trend for more of the volatile or reproductive development associated genes to be under positive We compared volatile and morphological divergence across the selection than other genes, the difference is not significant with a Caribbean Zamia clade to look for evidence of trait evolution proportion test (p = 0.1373). and our initial hypothesis was not related to the mutualism in the plant lineage and we found supported by these data. Gene annotation of the 30 positively a greater level of dissimilarity between Caribbean Zamia cone selected genes in the “volatile” bin described 21 molecular volatiles as compared to leaflet morphology (Figure 2). Our functions that fell into 7 categories. The highest percentage was analysis uncovered 48 volatile compounds across the entire uncategorized genes (41.46%, n = 17), followed by enzymatic Caribbean clade with an average of 15 compounds per individual, activity (24.39%, n = 10) and nucleotide binding (12.19%, n = 5) ranging from 9 to 26. Zamia integrifolia had the lowest average genes. The remaining categories are metal ion binding (9.76%, number of VOCs per species at 12.67 and Z. lucayana had the n = 4) transcription (4.88%, n = 2), ATP associated (4.88%, n = 2), highest average at 18.14. For leaflet morphology, ratios of leaflet and transport (2.44%, n = 1) genes (Supplementary Figure 6). length by width (L/W) range from 4.232 at the widest to 68.508 at the thinnest, with an average of 15.51. Zamia angustifolia had the thinnest leaflets with an average L/W ratio of 49.5 and Z. erosa DISCUSSION had the widest leaflets with an average ratio of 6.70, with all others falling within 9.5 (Z. lucayana) to 19.0 (Z. integrifolia). Our results provide evidence in support of co-evolutionary We used an analysis of similarity to determine if volatile reciprocal evolution in the Rhopalotria-Zamia pollination phenotypes are more dissimilar between taxa than leaflet mutualism. The relationship between two distantly related morphology. The ANOSIM R-value compares the mean of species pairs of Rhopalotria pollinators and their Zamia ranked dissimilarities between groups to the mean within groups. cycad hosts is characterized by a highly specific chemical Values below 0 indicated dissimilarities are greater within communication that is potentially evolving under positive groups than between groups, whereas those close to 1.0 suggest selection in the host plants, with phenotypes related to volatile dissimilarity between groups and those close to 0 indicate that production in the plants and perception in the insects displaying the distribution of values is even between and within groups. Our significant differences. We first identified a key trait underlying analysis showed a low level of dissimilarity between Caribbean the relationship and confirmed that chemical communication Zamia leaflet morphology using Bray-Curtis distance (R = 0.284) mediated the interaction by showing that the physiological and with a significance of 0.001. Conversely, we found an R-value behavioral responses of R. slossoni to a primary compound in closer to 1, and therefore, more dissimilar between groups for the volatiles expressed by their host plant (Figure 1) mirror plant volatile profiles (R = 0.4268, Bray-Curtis distance) with a the push-pull pollination mechanism previously described (Terry significance of 0.001. et al., 2007; Salzman et al., 2020). The behavioral response to methyl salicylate matches the behavioral response of R. furfuracea Searching for Selection on Zamia Volatile to its host plant compound 1,3-octadiene (Salzman et al., Associated Genes 2020). In the R. furfuracea/Z. furfuracea pollination system, To investigate the role of selection on divergence in plant Z. furfuracea initiates pollination by instigating the movement volatile traits, we sequenced de novo transcriptomes of pollen of R. furfuracea out of the host pollen cone through cyclical dehiscing cones of Caribbean Zamia and looked for evidence increases in the amount of 1,3-octadiene (Salzman et al., 2020). of positive selection in three gene “bins” determined by Here we found that methyl salicylate production undergoes a the differential expression pattern of genes compared to large increase in the pollen cone (Figure 1C) similarly to the methyl salicylate production or cone developmental stage. pattern observed with 1,3-octadiene in Z. furfuracea (Salzman Transcriptome assembly overall was of high quality with BUSCO et al., 2020). This burst in production of methyl salicylate occurs scores for 11 of the 12 species all showing 70.3–84.6% complete in the mid-afternoon and coincides with the time that R. slossoni BUSCO genes (Supplementary Figure 2). The single poor quality are most active (Tang, 1987). We infer from these results assembly was Zamia lucayana with only 8.6% complete BUSCO that that chemical communication and push-pull pollination genes, so we included this species in the phylogeny but not in is the usual mechanism underlying Rhopalotria and Zamia subsequent selection analysis. The maximum likelihood score associations and is likely to be found throughout the Cycadales for the phylogeny was −1,237,542.0107. The remaining 10 in- (Salzman et al., 2020). We then determined that these traits group Caribbean species transcriptomes had 2,487 orthologous are diverging between closely related taxa in both lineages. We

Frontiers in Plant Science| www.frontiersin.org 8 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 9

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

find that Rhopalotria pollinators respond to different chemicals nevertheless identify significant differences in volatile production in the volatiles released by their respective host plants, and and perception in these interactions, and point toward additional do not respond to an inappropriately matched chemical; they research that could be done to investigate how these interacting appear to be physiologically uncapable of perceiving their non- lineages may be affecting each other’s evolutionary trajectories. host plant. Plant volatiles differ more between species than Controversy still exists regarding the potential evolutionary their leaf morphology does (Figure 2), suggesting an increase trajectory of obligately coevolved mutualisms such as the push- in diversification in volatile phenotype compared to other pull pollinations of cycads (e.g., Vamosi et al., 2014; Day et al., morphological phenotypes, although without additional data 2016) and with further research, this system is likely to provide about trait evolution within and between these taxa of Zamia, it important insights. Divergence in chemical communication traits is difficult to conclude much about these relative differences. in both lineages has clear potential for contributing to lineage Disentangling the contribution of mutualistic species diversification, and while the fossil record supports a scenario interactions from the myriad of influences on diversification of plant coevolution with insect pollinators occurring before the remains elusive (Maron et al., 2019), yet fine scale investigations rise of angiosperms (Crepet, 1979) and even specifically involving of obligate mutualisms requiring a high degree of specialization cycads (Cai et al., 2018), the Rhopalotria/Zamia mutualism is between both partners offer some of the best potential insights likely to be much younger than these associations. Zamia is into the role of coevolution in diversification. The yucca/yucca thought to have arisen ∼ 85 Ma (Condamine et al., 2015; moth system is the only brood site pollination mutualism Calonje et al., 2019), about the same time as the earliest potential where these questions have been explicitly addressed so far, and divergence time for Rhopalotria (McKenna et al., 2009). However, research has shown that pollinating sister species of Tegeticula crown diversification of Zamia occurred much more recently at moths are driving divergence in floral traits related to pollination 22–9 Ma, with a rapid radiation occurring within the last ∼5 Ma (Godsoe et al., 2008) and contributing to reproductive isolation (Calonje et al., 2019). (Smith et al., 2009) in two varieties of Yucca brevifolia. Our To date, phylogenetic relationships between species of data suggest that chemical communication is likely to play a Rhopalotria have been based on a cladistic analysis of 89 central role in lineage diversification for both parties involved morphological, behavioral, and host characters, together with here as well. Selection can drastically alter plant volatile profiles 300 base pairs of 16s (Tang et al., 2018). This analysis over the course of just a few generations (Zu et al., 2016; described one species in Florida, Rhopalotria slossoni, one Gervasi and Schiestl, 2017; Ramos and Schiestl, 2020) and species, R. dimidiata, in the Bahamas and Cayman Islands, and the striking differences in volatile production and perception one in Jamaica, R. meerowi, with relationships between them shown here suggest that these traits are essential in securing unresolved. However, our initial analysis of 650 base pairs of CO1 species specificity. (unpublished) suggests greater diversity within these populations Volatile variation across populations has been described than has been previously appreciated. Larger scale phylogenetic in two cycad species, Encephalartos ghellinckii (Suinyuy and analysis is needed to determine the diversity of Rhopalotria and Johnson, 2018) and E. villosus (Suinyuy et al., 2013, 2015). related cycad pollinating members of the tribe Allocorynina in The former species has diverged in pollinator assemblage order to assess the potential co-diversification history of the two between populations while the later has not, but pollinators lineages. The interaction with specialized cycad herbivores on the do show preferences for volatiles from their host populations evolutionary trajectory of plant volatiles may also be important. both physiologically and behaviorally. Here, we show that two The interactive effects of pollinator attraction and herbivory can species of Rhopalotria weevils are physiologically specialized to have major and rapid effects on the evolution of plant traits perceive the active component of only their own host plant (Ramos and Schiestl, 2019, 2020). Methyl salicylate is involved VOCs, thereby exhibiting adaptation to their respective host in reproduction for Zamia integrifolia yet likely also has other plants. Scanning electron micrograph images of the antennae ecological functions as it is known to play a great many roles of different species of Rhopalotria show high variation in the in insect behavior from attraction (James, 2003) to deterrence distribution and density of sensory pockets (O’Brien and Tang, (Ninkovic et al., 2003) and even as an anti-aphrodisiac transferred 2015) that are likely to correspond with the physiological and in nuptial gifts (Andersson et al., 2000). Larvae of the lycaenid behavioral differences described here. Together, the evidence is butterfly, Eumaeus atala (Lepidoptera) larvae in Florida and the consistent with population divergence in volatile production or Caribbean are major pests of Zamia (Whitaker and Salzman, perception likely acting as an inhibitor to gene flow and thereby 2020) and while the chemical ecology of host plant localization facilitating diversification. We therefore looked for evidence or response to methyl salicylate remains unknown in E. atala, it of selection operating on genes related to volatile emission. likely further impacts the evolution and diversification of both Although a fifth of these genes were under positive selection, Zamia and Rhoplaotria lineages. we found no significant differences between the proportion of genes under selection in different categories (Supplementary Figure 5). It is possible that this was due to the limitation of DATA AVAILABILITY STATEMENT sample size of gene bins or because our process of binning volatile vs. reproductively related gene expression was not sufficient to The sequencing data has been deposited into a publicly accessible capture relevant differences. Overall, while we were unable to repository (link: https://dataverse.harvard.edu/dataverse/ confirm positive selection on volatile associated genes, our data CaribbeanZamia).

Frontiers in Plant Science| www.frontiersin.org 9 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 10

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

AUTHOR CONTRIBUTIONS collection possible. Steven Worthington, Melissa Whitaker, and Lisa DeGironimo provided advice on methods and/or comments SS, MC, DS, NP, and RH conceived of the project. SS performed on the manuscript. all field, laboratory and behavioral experiments, GCMS, and statistical analysis. DC performed EAD analysis. MC collected specimens in the field for morphological analysis. All authors SUPPLEMENTARY MATERIAL contributed to the article and approved the submitted version. The Supplementary Material for this article can be found FUNDING online at: https://www.frontiersin.org/articles/10.3389/fpls.2021. 639368/full#supplementary-material

SS was supported by the National Science Foundation Graduate Supplementary Figure 1 | Map of Caribbean Zamia cultivated at Montgomery Research Fellowship (NSF GRFP) and is now supported by the Botanical Center. NSF Postdoctoral Research Fellowship in Biology (1906333). Supplementary Figure 2 | BUSCO genes recovered in The Galaxy server that was used for some calculations assembled transcriptomes. is in part funded by Collaborative Research Centre 992 Medical Epigenetics (DFG grant SFB 992/1 2012) and German Supplementary Figure 3 | Volatile diversity across the Caribbean Zamia clade. Federal Ministry of Education and Research [BMBF grants Supplementary Figure 4 | Heatmap and dendrogram of Caribbean 031 A538A/A538C RBC, 031L0101B/031L0101C de.NBI-epi, Zamia volatiles. 031L0106 de.STAIR (de.NBI)]. Published with the aid of a grant Supplementary Figure 5 | Caribbean Zamia gene expression and from the Wetmore Colles fund. positive selection.

Supplementary Figure 6 | Functional analysis of volatile associated genes ACKNOWLEDGMENTS experiencing positive selection. Supplementary Table 1 | Accession information on Caribbean Zamia used in We thank Irene Terry for the use of her air sampling pumps and volatile analysis. guidance in volatile collection, Melissa Whitaker for assistance in Supplementary Table 2 | Accession information on Caribbean field work and Montgomery Botanical Center for access to their Zamia used for RNA. collections and the help of their dedicated and knowledgeable Supplementary Table 3 | Summary of transcriptome assemblies. staff, specifically Patrick Griffith, Xavier Gratacos, and Vickie Murphy. SS thanks her thesis committee, including Brian Farrell Supplementary Table 4 | Matrix occupancy stats for phylogenetic reconstruction. and David Haig, as well as Rory Maher for making data Supplementary Table 5 | Sequential Bonferroni corrected p-values for pitfall test.

REFERENCES Cai, C., Escalona, H. E., Li, L., Yin, Z., Huang, D., and Engel, M. S. (2018). Beetle pollination of cycads in the mesozoic. Curr. Biol. 28, 2806–2812. doi: 10.1016/ Afgan, E., Baker, D., Batut, B., van den Beek, M., Bouvier, D., Èech, j.cub.2018.06.036 M., et al. (2018). The Galaxy platform for accessible, reproducible and Calonje, M., Meerow, A. W., Griffith, M. P., Salas-Leiva, D., Vovides, A. P., Coiro, collaborative biomedical analyses: 2018 update. Nucleic Acids Res. 46, M., et al. (2019). A time-calibrated species tree phylogeny of the New World W537–W544. Cycad Genus Zamia L. (Zamiaceae. Cycadales). International Journal of Plant Althoff, D. M., Segraves, K. A., Smith, C. I., Leebens-Mack, J., and Pellmyr, O. Sciences. 180, 286–314. doi: 10.1086/702642 (2012). Geographic isolation trumps coevolution as a driver of yucca and yucca Calonje, M., Stevenson, D. W., and Osborne, R. (2020). The World List of Cycads. moth diversification. Mol. Phylogenet. Evol. 62, 898–906. doi: 10.1016/j.ympev. Cycads. 5, 77–119. 2011.11.024 Condamine, F. L., Nagalingum, N. S., Marshall, C. R., and Morlon, H. (2015). Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Miller, W., and Lipman, Origin and diversification of living cycads: a cautionary tale on the impact of the D. J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein branching process prior in Bayesian molecular dating. BMC Evolu. Biol. 15:65. database search programs. Nucleic Acids Res. 25, 3389–3402. doi: 10.1093/nar/ doi: 10.1186/s12862-015-0347-8 25.17.3389 Crepet, W. L. (1979). Insect pollination: a paleontological perspective. BioScience Andersson, J., Bor-Karlson, A. K., and Wiklund, C. (2000). Sexual cooperation and 29, 102–108. doi: 10.2307/1307746 conflict in butterflies: a male-transferred anti-aphrodisiac reduces harassment Crook, D. J., Khrimian, A., Francese, J. A., Fraser, I., Poland, T. M., Sawyer, of recently mated females. Proc. R. Soc. London B 267, 1271–1275. doi: 10.1098/ A. J., et al. (2008). development of a host-based semiochemical lure for rspb.2000.1138 trapping emerald ash borer Agrilus planipennis (Coleoptera: Buprestidae). Ashburner, M., Ball, C. A., Blake, J. A., Botstein, D., Butler, H., Cherry, J. M., et al. Environmental Entomology. 37, 356–365. doi: 10.1603/0046-225x(2008)37[356: (2000). Gene ontology: tool for the unification of biology. Nat. Genet. 25, 25–29. doahsl]2.0.co;2 Bolger, A. M., Lohse, M., and Usadel, B. (2014). Trimmomatic: a flexible trimmer Day, E. H., Hua, X., and Bromham, L. (2016). Is specialization an evolutionary for Illumina Sequence Data. Bioinformatics. 30, 2114–2120. doi: 10.1093/ dead end? Testing for differences in speciation, extinction and trait transition bioinformatics/btu170 rates across diverse phylogenies of specialists and generalists. . Evol. Biol. 29, Bryant, D. M., Johnson, K., DiTommaso, T., Tickle, T., Couger, M. B., Payzin- 1257–1267. doi: 10.1111/jeb.12867 Dogru, D., et al. (2017). A tissue-mapped axolotl de novo transcriptome enables Dunn, C. W., Hejnol, A., Matus, D. Q., Pang, K., Browne, W. E., Smith, S. A., et al. identification of limb regeneration factors. Cell Rep. 18, 762–776. doi: 10.1016/ (2008). Broad phylogenomic sampling improves resolution of the animal tree j.celrep.2016.12.063 of life. Nature 452, 745–749. doi: 10.1038/nature06614

Frontiers in Plant Science| www.frontiersin.org 10 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 11

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

Dunn, C. W., Howison, M., and Zapata, F. (2013). Agalma: an automated McKenna, D. D., Sequeira, A. S., Marvaldi, A. E., and Farrell, B. D. (2009). phylogenomics workflow. BMC Bioinformatics. 14:330. doi: 10.1186/1471- Temporal lags and overlap in the diversification of weevils and flowering plants. 2105-14-330 PNAS 106, 7083–7088. doi: 10.1073/pnas.0810618106 Eckenwalder, J. E. (1980). Taxonomy of the West Indian cycads. J. Arnold Mound, L. A., and Terry, I. (2001). Thrips pollination of the central Australian Arboretum 61, 701–722. doi: 10.5962/bhl.part.8542 cycad Macrozamia macdonnellii (Cycadales). Int. J. Plant Sci. 162, 147–154. Enright, A. J., Van Dongen, S., and Ouzounis, C. A. (2002). An efficient algorithm doi: 10.1086/317899 for large-scale detection of protein families. Nucleic Acids Res. 30, 1575–1584. Nguyen, L.-T., Schmidt, H. A., von Haeseler, A., and Minh, B. Q. (2015). IQ-TREE: doi: 10.1093/nar/30.7.1575 a fast and effective stochastic algorithm for estimating maximum likelihood Gervasi, D. D. L., and Schiestl, F. P. (2017). Real-time divergent evolution in plants phylogenies. Mol. Biol. Evol. 32, 268–274. doi: 10.1093/molbev/msu300 driven by pollinators. Nat. Commun. 8:14691. Nielsen, R., and Yang, Z. (1998). Likelihood models for detecting positively selected Godsoe, W., Yoder, J. B., Smith, C. I., and Pellmyr, O. (2008). Coevolution and amino acid sites and applications to the HIV-1 envelope gene. Genetics 148, divergence in the Joshua tree/yucca moth mutualism. Am Nat. 171, 816–823. 929–936. doi: 10.1093/genetics/148.3.929 doi: 10.1086/587757 Ninkovic, V., Ahmed, E., Glinwood, R., and Pettersson, J. (2003). Effects of two Grabherr, M. G., Haas, B. J., Yassour, M., Levin, J. Z., Thompson, D. A., Amit, I., types of semiochemical on population development of the bird cherry oat aphid et al. (2011). Full-length transcriptome assembly from RNA-seq data without a Rhopalosiphum padi in a barley crop. Agric. Forest Entomol. 5, 27–33. reference genome. Nat. Biotechnol. 29, 644–652. doi: 10.1038/nbt.1883 Norstog, K. J., and Fawcett, P. K. S. (1989). Insect-cycad symbiosis and its relation Griffith, M. P., Calonje, M. A., Stevenson, D. W., Husby, C. E., and Little, D. P. to the pollination of Zamia furfuracea (Zamiaceae) by Rhopalotria mollis (2012). Time, place, and relationships: cycad phenology in a phylogenetic and (Curculionidae). Ame. J. Bot. 76, 1380–1394. biogeographic context. Mem. N. Y. Bot. Garden 106, 59–81. Norstog, K. J., Fawcett, P. K. S., and Vovides, A. P. (1992). Beetle pollination of two Haas, B. J., Papanicolaou, A., Yassour, M., Grabherr, M., Blood, P. D., Bowden, J., species of Zamia: Evolutionary and ecological considerations. Paleobotanist 41, et al. (2013). De novo transcript sequence reconstruction from RNA-seq using 149–158. the Trinity platform for reference generation and analysis. Nat. Protocols 8, Norstog, K. J., Stevenson, D. W., and Niklas, K. J. (1986). The role of beetles in 1494–1512. doi: 10.1038/nprot.2013.084 the pollination of Zamia furfuracea L. fil.(Zamiaceae). Biotropica 18, 300–306. Hembry, D. H., Kawakita, A., Gurr, N. E., Schmaedick, M. A., Baldwin, B. G., and doi: 10.2307/2388573 Gillespie, R. G. (2013). Non-congruent colonizations and diversification in a O’Brien, C. W., and Tang, W. (2015). Revision of the New World cycad weevils coevolving pollination mutualism on oceanic islands. Proc R Soc B: Biol. Sci. of the subtribe Allocorynina, with description of two new genera and three 280:20130361. doi: 10.1098/rspb.2013.0361 new subgenera (Coleoptera: Belidae: Oxycoryninae). Zootaxa 3970, 1–87. doi: Herre, E. A., Jandér, K. C., and Machado, C. A. (2008). Evolutionary ecology of figs 10.11646/zootaxa.3970.1.1 and their associates: recent progress and outstanding puzzles. Annu Rev Ecol. Okamoto, T., Kawakita, A., and Kato, M. (2007). Interspecific variation of Evol. Syst. 39, 439–458. doi: 10.1146/annurev.ecolsys.37.091305.110232 floral scent composition in Glochidion and its association with host-specific Holland, J. N., and Fleming, T. H. (1999). Mutualistic interactions between Upiga pollinating seed parasite (Epicephala). J. Chem. Ecol. 33, 1065–1081. doi: 10. virescens (Pyralidae), a pollinating seed consumer, and Lophocereus schottii 1007/s10886-007-9287-0 (Cactaceae). Ecology 80, 2074–2084. doi: 10.2307/176679 Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, James, D. G. (2003). Field evaluation of herbivore-induced plant volatiles as D., et al. (2019). vegan: Community Ecology Package. R package version 2.5- attractants for beneficial insects: methyl salicylate and the green lacewing, 6. Available online at: https://CRAN.R-project.org/package=vegan (accessed Chrysopa nigricornis. J. Chem. Ecol. 29, 1601–1609. January, 2021). Janzen, D. H. (1980). When is it coevolution? Evolution 34, 611–612. doi: 10.2307/ Pellmyr, O. (2003). Yuccas, yucca moths, and coevolution: a review. Annu. Missouri 2408229 Bot. Garden 90, 30–55. Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A., and Jermiin, Pellmyr, O., Tang, W., Groth, I., Bergström, G., and Thien, L. B. (1991). Cycad L. S. (2017). ModelFinder: fast model selection for accurate phylogenetic cone and angiosperm floral volatiles: inferences for the evolution of insect estimates. Nat. Methods 14, 587–589. doi: 10.1038/nmeth.4285 pollination. Biochem. Syst. Ecol. 19, 623–627. doi: 10.1016/0305-1978(91) Katoh, K., and Standley, D. M. (2013). MAFFT multiple sequence alignment 90078-e software version 7: improvements in performance and usability. Mol. Biol. Evol. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., and R Core Team (2020). nlme: Linear 30, 772–780. doi: 10.1093/molbev/mst010 and Nonlinear Mixed Effects Models. R package version 3.1-148. Available online Kawakita, A. (2010). Evolution of obligate pollination mutualism in the tribe at: https://CRAN.R-project.org/package=nlme (accessed January, 2021). Phyllantheae. Plant Species Biol. 25, 3–19. doi: 10.1111/j.1442-1984.2009. Price, M. N., Dehal, P. S., and Arkin, A. P. (2010). FastTree 2—approximately 00266.x maximum-likelihood trees for large alignments. PLoS One 5:e9490. doi: 10. Kovats, E. S. (1965). “Gas chromotographic characterization of organic substances 1371/journal.pone.0009490 in the retention index system,” in Advances in Chromatography, eds J. C. R Core Team (2020). R: A Language and Environment for Statistical Computing. Giddings and R. A. Keller (New York NY: Dekker), 229–247. Vienna: R Foundation for Statistical Computing. Leng, N., Dawson, J. A., Thomson, J. A., Ruotti, V., Rissman, A. I, Smits, Ramos, S. E., and Schiestl, F. P. (2019). Rapid plant evolution driven by the B. M. G., et al. (2013). EBSeq: An empirical Bayes hierarchical model for interaction of pollination and herbivory. Science 364, 193–196. doi: 10.1126/ inference in RNA-seq experiments. Bioinformatics 29, 1035–1043. doi: 10.1093/ science.aav6962 bioinformatics/btt087 Ramos, S. E., and Schiestl, F. P. (2020). Evolution of floral fragrance is Leng, N., and Kendziorski, C. (2020). EBSeq: An R Package for Gene and Isoform compromised by herbivory. Front. Ecol. Evol. 8:30. doi: 10.3389/fevo.2020. Differential Expression Analysis of RNA-seq Data. R Package Version 1.28.0. 00030 Li, B., and Dewey, C. N. (2011). RSEM: accurate transcript quantification for RNA- Ran, J. H., Shen, T. T., Wang, M. M., and Wang, X. Q. (2018). Phylogenomics Seq data with or without a reference genome. BMC Bioinformatics. 12:323. resolves the deep phylogeny of seed plants and indicates partial convergent doi: 10.1186/1471-2105-12-323 or homoplastic evolution between Gnetales and angiosperms. Proc. R. Soc. B. Li, W., and Godzik, A. (2006). Cd-hit: a fast program for clustering and comparing 285:20181012. doi: 10.1098/rspb.2018.1012 large sets of protein or nucleotide sequences. Bioinformatics 22, 1658–1659. Rueden, C. T., Schindelin, J., Hiner, M. C., DeZonia, B. D., Walter, A. E., doi: 10.1093/bioinformatics/btl158 Arena, E. T., et al. (2017). ImageJ2: ImageJ for the next generation of Löytynoja, A., and Goldman, N. (2008). Phylogeny-aware gap placement prevents scientific image data. BMC Bioinformatics. 18:529. doi: 10.1186/s12859-017- errors in sequence alignment and evolutionary analysis. Science 320, 1632–1635. 1934-z doi: 10.1126/science.1158395 Salzman, S., Crook, D., Crall, J. D., Hopkins, R., and Pierce, N. E. (2020). An Maron, J. L., Agrawal, A. A., and Schemske, D. W. (2019). Plant-herbivore ancient push-pull pollination mechanism in cycads. Sci. Ad. 6:eaay6169. doi: coevolution and plant speciation. Ecology 100:e02704. 10.1126/sciadv.aay6169

Frontiers in Plant Science| www.frontiersin.org 11 April 2021| Volume 12| Article 639368 fpls-12-639368 April 30, 2021 Time: 12:6 # 12

Salzman et al. Chemical Communication in Cycad-Weevil Pollination Mutualism

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Terry, I., Walter, G. H., Donaldson, J. S., Snow, E., Forster, P. I., and Machin, P. J. et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat. (2005). Pollination of Australian Macrozamia cycads (Zamiaceae): Effectiveness Methods 9, 676–682. doi: 10.1038/nmeth.2019 and behavior of specialist vectors in a dependent mutualism. Am. J. Bot. 92, Schultz, A. J., and Sackton, T. B. (2019). Immune genes are hotspots of shared 931–940. doi: 10.3732/ajb.92.6.931 positive selection across birds and mammals. eLife 2019, e41815. Terry, I., Walter, G. H., Moore, C., and Roemer, C. (2007). Odor-mediated Seppey, M., Manni, M., and Zdobnov, E. M. (2019). “BUSCO: assessing genome push-pull pollination in cycads. Science 318:70. doi: 10.1126/science. assembly and annotation completeness,” in Gene Prediction. Methods in 1145147 Molecular Biology, Vol. 1962, ed. M. Kollmar (New York, NY: Humana), Gene Ontology Consortium (2021). The Gene Ontology resource enriching a Gold 227–245. doi: 10.1007/978-1-4939-9173-0_14 mine. Nucleic Acid Res. 49, D325–D334. Smith, C. I., Drummond, C. S., Godsoe, W., Yoder, J. B., and Pellmyr, O. Vamosi, J. C., Armbruster, W. S., and Renner, S. S. (2014). Evolutionary ecology of (2009). Host specificity and reproductive success of yucca moths (Tegeticula specialization: insights from phylogenetic analysis. Proc. R. Soc. London B Biol. spp. Lepidoptera: Prodoxidae) mirrors patterns of gene flow between host Sci. 281:20142004. doi: 10.1098/rspb.2014.2004 plant varieties of the Joshua tree (Yucca brevifolia: Agavaceae). Mol. Ecol. 18, van Den Dool, H., and Kratz, P. D. (1963). A generalization of the retention 5218–5229. doi: 10.1111/j.1365-294x.2009.04428.x index system including linear temperature programmed gas-liquid partition Smith, S. A., and Dunn, C. W. (2008). Phyutility: a phyloinformatics tool for chromatography. J Chromatogr. 11, 463–471. doi: 10.1016/s0021-9673(01) trees, alignments and molecular data. Bioinformatics 24, 715–716. doi: 10.1093/ 80947-x bioinformatics/btm619 Van Dongen, S., and Abreu-Goodger, C. (2012). Using MCL to extract clusters Smith, S. A., Wilson, N. G., Goetz, F. E., Feehery, C., Andrade, S. C. S., Rouse, from networks. Bactl Mol Networks Methods Protocols 804, 281–295. doi: G. W., et al. (2011). Resolving the evolutionary relationships of molluscs with 10.1007/978-1-61779-361-5_15 phylogenomic tools. Nature 480, 364–367. doi: 10.1038/nature10526 Whitaker, M. R. L., and Salzman, S. (2020). Ecology and evolution of Stamatakis, A. (2014). RAxML version 8: a tool for phylogenetic analysis and cycad-feeding Lepidoptera. Ecol Lett. 23, 1862–1877. doi: 10.1111/ele. post-analysis of large phylogenies. Bioinformatics 30, 1312–1313. doi: 10.1093/ 13581 bioinformatics/btu033 Wong, W. S., Yang, Z., Goldman, N., and Nielsen, R. (2004). Accuracy and power of Stevenson, D. W., Norstog, K. J., and Fawcett, P. K. S. (1998). “Pollination biology statistical methods for detecting adaptive evolution in protein coding sequences of cycads,” in Reproductive Biology, eds S. J. Owens and P. J. Rudall (London: and for identifying positively selected sites. Genetics 168, 1041–1051. doi: Kew Gardens), 277–294. 10.1534/genetics.104.031153 Strimmer, K. (2008). Fdrtool: a versatile R packages for estimating local and tail Yang, Y., and Smith, S. A. (2014). Orthology inference in nonmodel organisms area-based false discovery rates. Bioinformatics. 24, 1461–1462. doi: 10.1093/ using transcriptomes and low-coverage genomes: improving accuracy and bioinformatics/btn209 matrix occupancy for phylogenomics. Mol. Biol. Evol. 31, 3081–3092. doi: Suinyuy, T. N., Donaldson, J. S., and Johnson, S. D. (2013). Patterns of odour 10.1093/molbev/msu245 emission, thermogenesis and pollinator activity in cones of an African cycad: Yang, Z. (2007). PAML 4: phylogenetic analysis by maximum likelihood. Mol. Biol. what mechanisms apply? Ann. Bot. 112, 891–902. doi: 10.1093/aob/mct159 Evol. 24, 1586–1591. doi: 10.1093/molbev/msm088 Suinyuy, T. N., Donaldson, J. S., and Johnson, S. D. (2015). Geographical matching Yang, Z., Nielsen, R., Goldman, N., and Pedersen, A. M. (2000). Codon-substitution of volatile signals and pollinator olfactory responses in a cycad brood-site models for heterogeneous selection pressure at amino acid sites. Genetics 155, mutualism. Proc. R. Soc. B Bio.l Sci. 282:20152053. doi: 10.1098/rspb.2015.2053 431–449. Suinyuy, T. N., and Johnson, S. D. (2018). Geographic variation in cone volatiles Zu, P., Blanckenhorn, W. U., and Schiestl, F. P. (2016). Heritability of floral and pollinators in the thermogenic African cycad Encephalartos ghellinckii volatiles and pleiotropic responses to artificial selection in Brassica rapa. New Lem. Plant Biol. 20, 579–590. doi: 10.1111/plb.12685 Phytol. 209, 1208–1219. doi: 10.1111/nph.13652 Tang, W. (1987). Insect pollination in the cycad Zamia pumila (Zamiaceae). Am. J. Bot. 74, 90–99. doi: 10.2307/2444334 Conflict of Interest: The authors declare that the research was conducted in the Tang, W., Xu, G., O’Brien, C., Calonje, M., Franz, N., Johnston, M. A., et al. (2018). absence of any commercial or financial relationships that could be construed as a Molecular and Morphological phylogenetic analyses of new world cycad beetles: potential conflict of interest. what they reveal about cycad evolution in the New World. Diversity 10:38. doi: 10.3390/d10020038 Copyright © 2021 Salzman, Crook, Calonje, Stevenson, Pierce and Hopkins. This Terry, I. (2001). Thrips and weevils as dual specialist pollinators of the Australian is an open-access article distributed under the terms of the Creative Commons cycad Macrozamia communis (Zamiaceae). Int. J Plant Sci. 162, 1293–1305. Attribution License (CC BY). The use, distribution or reproduction in other forums doi: 10.1086/321929 is permitted, provided the original author(s) and the copyright owner(s) are credited Terry, I., Tang, W., Taylor, A. S., Donaldson, J. S., Singh, R., Vovides, A. P., et al. and that the original publication in this journal is cited, in accordance with accepted (2012). An overview of cycad pollination studies. Mem N. Y. Bot. Garden 106, academic practice. No use, distribution or reproduction is permitted which does not 352–394. comply with these terms.

Frontiers in Plant Science| www.frontiersin.org 12 April 2021| Volume 12| Article 639368